Mystery: examining a crime scene

Bad Writing Habits

Detects common prose quality anti-patterns in AI-generated creative writing, including passive voice, past progressive overuse, weak dialogue tags, filter words, purple prose, cliches, AI-ism words/adverbs/names, and more.

Creative Writing Hallucination

Performance Score Distribution (Top 20)

Click a model name to view its detail page.

	Score
GPT-5.4 (Reasoning)	92%
GPT-5.4	92%
GPT-5.4 (Reasoning, Low)	91%
GPT-5.4 Mini	89%
GPT-5.4 Mini (Reasoning)	89%
GPT-5.4 Mini (Reasoning, Low)	87%
GPT-5	85%
GPT-5.1	85%
Rocinante 12B	85%
Claude Sonnet 4.5	85%
Qwen 3.5 397B A17B	84%
Mistral Small 4 (Reasoning)	83%
Qwen 3 32B	83%
Claude Opus 4	83%
Grok 4.20 (Beta, Reasoning)	83%
Mistral Large 3	82%
Gemini 3.1 Pro (Preview)	82%
Hermes 3 405B	82%
MiniMax M2.5	82%
Claude Sonnet 4.6	82%

	Score	Cost	Time
Rocinante 12B	85%	$0.0006	25.4s
GPT-5.4 Mini (Reasoning, Low)	87%	$0.016	18.8s
GPT-5.4 Mini	89%	$0.017	19.5s
Mistral Small 4 (Reasoning)	83%	$0.0022	30.7s
Qwen 3 32B	83%	$0.0012	39.7s
Mistral Small Creative	81%	$0.0006	9.8s
GPT-5.4 Nano	80%	$0.0057	20.4s
Mistral NeMO	81%	$0.0003	11.1s
Mistral Large 3	82%	$0.0026	28.4s
Mistral Small 4	79%	$0.0011	18.0s
LFM2 24B	78%	$0.0002	33.2s
Ministral 3 14B	79%	$0.0005	12.5s
GPT-5.4 Nano (Reasoning, Low)	81%	$0.0055	20.4s
GPT-5.4 Mini (Reasoning)	89%	$0.031	37.4s
DeepSeek V3 (2025-03-24)	80%	$0.0016	33.7s
Qwen 3.5 Flash	81%	$0.0019	36.5s
o4 Mini	81%	$0.014	22.3s
Hermes 3 405B	82%	$0.0020	59.2s
Grok 4.1 Fast	80%	$0.0016	40.6s
Gemma 3 12B	79%	$0.0004	37.1s

	Score	Consistency	Stability
GPT-5.4 (Reasoning)	92%	98%	91%
GPT-5.4	92%	96%	88%
GPT-5.4 (Reasoning, Low)	91%	94%	87%
GPT-5.4 Mini (Reasoning, Low)	87%	96%	84%
GPT-5.4 Mini (Reasoning)	89%	95%	84%
GPT-5.1	85%	98%	83%
GPT-5	85%	96%	83%
Claude Sonnet 4.5	85%	96%	81%
Qwen 3.5 397B A17B	84%	95%	80%
Qwen 3 32B	83%	97%	80%
GPT-5.4 Mini	89%	92%	80%
Claude Opus 4	83%	94%	79%
Mistral Small 4 (Reasoning)	83%	94%	78%
o4 Mini	81%	94%	77%
Writer: Palmyra X5	81%	97%	77%
GPT-5.4 Nano (Reasoning, Low)	81%	95%	77%
GPT-5.4 Nano	80%	94%	77%
Grok 4	79%	98%	77%
Rocinante 12B	85%	88%	77%
Qwen3 235B A22B Instruct 2507	81%	95%	77%

	Score	Cost	Speed	Stability
GPT-5.4 Mini (Reasoning, Low)	87%	$0.016	18.8s	84%
GPT-5.4 Mini	89%	$0.017	19.5s	80%
GPT-5.4 Mini (Reasoning)	89%	$0.031	37.4s	84%
Rocinante 12B	85%	$0.0006	25.4s	77%
Qwen 3 32B	83%	$0.0012	39.7s	80%
GPT-5.4	92%	$0.057	1.6m	88%
Mistral Small 4 (Reasoning)	83%	$0.0022	30.7s	78%
Claude Sonnet 4.5	85%	$0.029	35.7s	81%
Mistral Large 3	82%	$0.0026	28.4s	76%
GPT-5.4 (Reasoning, Low)	91%	$0.060	1.6m	87%
Mistral Small Creative	81%	$0.0006	9.8s	75%
Mistral NeMO	81%	$0.0003	11.1s	75%
GPT-5.4 Nano (Reasoning, Low)	81%	$0.0055	20.4s	77%
GPT-5.4 Nano	80%	$0.0057	20.4s	77%
Writer: Palmyra X5	81%	$0.011	24.2s	77%
o4 Mini	81%	$0.014	22.3s	77%
Ministral 3 14B	79%	$0.0005	12.5s	75%
Qwen 3.5 Flash	81%	$0.0019	36.5s	75%
Mistral Small 4	79%	$0.0011	18.0s	75%
MiniMax M2.5	82%	$0.0030	1.0m	76%

Rank	Model	Avg. Cost	Avg. Time	Stability	# 1	# 2	# 3	# 4	# 5	Total
51	GPT-5.4 (Reasoning)	$0.100	3.1m	91%	94	93	93	91	91	92%
6	GPT-5.4	$0.057	1.6m	88%	95	92	92	90	89	92%
10	GPT-5.4 (Reasoning, Low)	$0.060	1.6m	87%	93	93	92	89	85	91%
2	GPT-5.4 Mini	$0.017	19.5s	80%	95	90	87	86	85	89%
3	GPT-5.4 Mini (Reasoning)	$0.031	37.4s	84%	93	89	88	87	86	89%
1	GPT-5.4 Mini (Reasoning, Low)	$0.016	18.8s	84%	89	88	87	87	84	87%
88	GPT-5	$0.073	3.0m	83%	87	87	86	83	83	85%
43	GPT-5.1	$0.063	1.5m	83%	86	86	85	85	83	85%
4	Rocinante 12B	$0.0006	25.4s	77%	92	89	87	79	76	85%
8	Claude Sonnet 4.5	$0.029	35.7s	81%	87	85	84	84	82	85%
47	Qwen 3.5 397B A17B	$0.030	2.3m	80%	86	85	85	85	79	84%
7	Mistral Small 4 (Reasoning)	$0.0022	30.7s	78%	87	86	83	82	78	83%
5	Qwen 3 32B	$0.0012	39.7s	80%	85	84	83	83	80	83%
111	Claude Opus 4	$0.147	1.0m	79%	85	85	84	83	77	83%
31	Grok 4.20 (Beta, Reasoning)	$0.036	33.9s	77%	86	86	83	82	76	83%
9	Mistral Large 3	$0.0026	28.4s	76%	88	82	81	81	80	82%
108	Gemini 3.1 Pro (Preview)	$0.083	1.5m	71%	91	88	80	78	75	82%
24	Hermes 3 405B	$0.0020	59.2s	74%	88	86	82	78	76	82%
20	MiniMax M2.5	$0.0030	1.0m	76%	87	83	83	81	75	82%
40	Claude Sonnet 4.6	$0.031	48.5s	77%	87	83	83	77	77	82%
22	Qwen3 235B A22B Instruct 2507	$0.0011	1.1m	77%	85	84	81	79	78	81%
18	Qwen 3.5 Flash	$0.0019	36.5s	75%	88	81	80	79	78	81%
33	Mistral Large 2	$0.011	28.7s	72%	89	84	79	78	76	81%
15	Writer: Palmyra X5	$0.011	24.2s	77%	83	83	80	80	79	81%
16	o4 Mini	$0.014	22.3s	77%	83	83	82	79	76	81%
12	Mistral NeMO	$0.0003	11.1s	75%	87	80	80	80	77	81%
11	Mistral Small Creative	$0.0006	9.8s	75%	86	81	80	78	78	81%
103	Claude Opus 4.6 (Reasoning)	$0.086	1.4m	76%	84	84	81	79	76	81%
13	GPT-5.4 Nano (Reasoning, Low)	$0.0055	20.4s	77%	84	81	81	79	77	81%
21	Grok 4.1 Fast	$0.0016	40.6s	75%	84	83	80	79	75	80%
14	GPT-5.4 Nano	$0.0057	20.4s	77%	83	83	82	79	75	80%
42	o4 Mini High	$0.023	38.2s	74%	86	82	80	78	76	80%
37	Z.AI GLM 5 Turbo	$0.0081	37.9s	72%	87	83	80	77	74	80%
25	GPT-5 Mini	$0.0099	40.3s	76%	84	79	79	79	79	80%
41	Mistral Large	$0.013	35.0s	73%	87	80	80	76	75	80%
96	GPT-5.2	$0.060	1.6m	75%	84	80	79	78	78	80%
27	DeepSeek V3 (2025-03-24)	$0.0016	33.7s	74%	83	83	80	80	72	80%
48	GPT-4.1	$0.018	40.9s	73%	84	83	79	76	75	79%
32	Gemma 3 27B	$0.0007	45.4s	73%	85	83	80	77	73	79%
19	Mistral Small 4	$0.0011	18.0s	75%	84	81	80	78	74	79%
35	Stealth: Healer Alpha	$0.0000	24.6s	71%	85	84	78	77	72	79%
55	Ministral 3 8B	$0.0023	1.5m	73%	86	81	80	75	75	79%
29	Gemini 3.1 Flash Lite (Preview)	$0.0026	8.1s	71%	86	82	77	76	76	79%
17	Ministral 3 14B	$0.0005	12.5s	75%	82	81	80	79	74	79%
71	Gemini 2.5 Pro	$0.040	41.1s	72%	84	82	78	77	74	79%
26	Mistral Medium 3.1	$0.0045	34.0s	76%	82	81	81	77	74	79%
50	Qwen 3.5 9B	$0.0008	1.2m	72%	85	80	78	75	75	79%
30	ByteDance Seed 1.6 Flash	$0.0014	30.1s	73%	84	80	78	77	75	79%
39	DeepSeek-V2 Chat	$0.0016	36.9s	72%	84	81	80	79	70	79%
23	Gemma 3 12B	$0.0004	37.1s	76%	81	81	80	76	75	79%
75	Cohere Command R+ (Aug. 2024)	$0.016	1.3m	71%	84	84	79	75	71	79%
86	Grok 4	$0.045	1.7m	77%	80	79	78	78	77	79%
28	GPT-5.4 Nano (Reasoning)	$0.0060	22.2s	75%	82	80	79	77	74	78%
44	Gemini 2.5 Flash (Reasoning)	$0.013	24.5s	72%	83	81	80	79	69	78%
68	Claude 3.5 Sonnet	$0.044	37.5s	75%	81	80	78	77	76	78%
45	Qwen 3.5 122B	$0.013	36.8s	74%	83	81	81	74	72	78%
56	Grok 4.20 (Beta)	$0.015	13.1s	69%	88	78	78	75	72	78%
34	LFM2 24B	$0.0002	33.2s	73%	83	82	81	74	71	78%
70	Claude 3.7 Sonnet	$0.037	46.9s	73%	83	79	79	78	72	78%
99	Claude Opus 4.5	$0.067	56.1s	73%	82	80	77	77	75	78%
97	ByteDance Seed 2.0 Lite	$0.012	2.1m	68%	89	80	79	72	70	78%
54	MiniMax M2.7	$0.0033	1.3m	74%	82	80	78	76	75	78%
106	Claude Opus 4.6	$0.077	1.3m	75%	81	80	78	76	74	78%
59	Z.AI GLM 5	$0.0079	1.1m	73%	83	77	77	77	77	78%
58	GPT-4o, May 13th (temp=0)	$0.028	19.0s	73%	82	80	77	76	74	78%
38	Gemini 2.5 Flash	$0.0057	11.6s	73%	81	78	76	76	74	77%
36	Gemini 2.5 Flash Lite	$0.0009	11.8s	72%	80	79	77	76	72	77%
57	Claude 3 Haiku	$0.0020	15.6s	68%	85	77	75	73	72	76%
62	Grok 4 Fast	$0.0016	27.2s	68%	82	82	75	74	69	76%
78	Qwen 3.5 27B	$0.0097	57.2s	69%	84	77	75	73	72	76%
81	Qwen 3.5 35B	$0.011	37.9s	66%	86	77	74	73	72	76%
60	Llama 3.1 Nemotron 70B	$0.0023	32.3s	70%	83	78	76	72	72	76%
118	ByteDance Seed 2.0 Mini	$0.0049	5.4m	75%	77	77	76	76	75	76%
53	Llama 3.1 70B	$0.0007	33.1s	71%	80	78	75	74	74	76%
52	GPT-4o Mini (temp=1)	$0.0011	25.2s	71%	81	75	75	75	74	76%
113	DeepSeek V3.2	$0.0011	3.3m	67%	84	78	75	72	70	76%
49	Gemini 2.5 Flash Lite (Reasoning)	$0.0025	26.5s	72%	79	78	75	74	74	76%
61	Arcee AI: Trinity Large (Preview)	$0.0000	42.7s	70%	80	78	75	73	72	76%
93	MoonshotAI: Kimi K2.5	$0.012	1.0m	66%	86	77	76	70	68	75%
80	GPT-4o, May 13th (temp=1)	$0.030	15.6s	70%	82	76	75	72	71	75%
104	Claude Sonnet 4.6 (Reasoning)	$0.043	1.1m	70%	81	77	77	74	67	75%
89	Hermes 3 70B	$0.0006	31.1s	62%	84	82	71	71	69	75%
100	Z.AI GLM 4.7	$0.0098	1.9m	69%	80	77	74	73	72	75%
110	ByteDance Seed 1.6	$0.016	3.1m	72%	78	75	74	74	74	75%
87	Aion 2.0	$0.0051	1.2m	68%	81	79	74	72	69	75%
46	GPT-4o Mini (temp=0)	$0.0010	24.7s	73%	77	76	75	74	73	75%
64	DeepSeek V3 (2024-12-26)	$0.0017	40.7s	71%	78	77	75	74	70	75%
83	Ministral 3B	$0.0001	1.1m	68%	82	76	73	72	71	75%
74	Ministral 3 3B	$0.0015	55.5s	70%	80	77	75	73	70	75%
63	Qwen 2.5 72B	$0.0006	27.7s	70%	79	79	75	71	71	75%
94	Z.AI GLM 4.7 Flash	$0.0020	1.3m	66%	84	73	73	72	70	75%
107	Gemini 3 Pro (Preview)	$0.049	49.7s	69%	79	78	74	71	69	74%
101	Claude Sonnet 4	$0.028	43.4s	66%	79	79	74	73	65	74%
76	Stealth: Hunter Alpha	$0.0000	46.8s	69%	80	77	76	69	68	74%
92	DeepSeek V3.1	$0.0014	1.7m	70%	78	76	75	71	70	74%
69	GPT-4o, Aug. 6th (temp=0)	$0.016	29.9s	73%	75	74	74	74	73	74%
98	Z.AI GLM 4.6	$0.0062	1.4m	67%	78	77	72	72	70	74%
82	GPT-4o, Aug. 6th (temp=1)	$0.016	36.9s	70%	79	75	75	71	69	74%
66	GPT-4.1 Mini	$0.0027	16.6s	69%	78	76	74	70	70	74%
85	Claude Haiku 4.5	$0.0093	20.2s	66%	80	79	73	69	67	74%
65	Gemma 3 4B	$0.0002	18.9s	70%	76	75	74	72	70	73%
72	WizardLM 2 8x22b	$0.0016	24.6s	69%	78	77	76	69	66	73%
84	Qwen 3.5 Plus (2026-02-15)	$0.0051	25.8s	67%	79	74	72	71	70	73%
112	Llama 3.1 8B	$0.0003	2.0m	63%	82	77	70	69	68	73%
73	Ministral 8B	$0.0002	9.9s	67%	80	77	76	68	63	73%
90	Z.AI GLM 4.5	$0.0045	37.2s	67%	79	73	73	72	66	73%
67	GPT-4.1 Nano	$0.0006	10.5s	70%	75	74	74	71	68	72%
79	Gemini 3 Flash (Preview)	$0.0067	17.5s	69%	74	72	72	71	68	72%
77	Arcee AI: Trinity Mini	$0.0003	9.0s	67%	77	75	73	68	66	72%
95	Gemini 3 Flash (Preview, Reasoning)	$0.010	25.1s	66%	78	73	72	68	66	72%
91	Claude 3.5 Haiku	$0.0027	12.5s	65%	77	75	74	67	61	71%
109	Nemotron 3 Super	$0.0000	1.3m	65%	74	70	68	68	67	69%
116	Mistral Small 3.2 24B	$0.0034	1.8m	62%	77	70	69	68	60	69%
102	Inception Mercury 2	$0.0031	7.9s	65%	68	68	67	66	65	67%
115	GPT-5 Nano	$0.0042	1.4m	63%	71	67	66	65	64	67%
117	Nemotron 3 Nano	$0.0007	42.1s	56%	77	69	63	63	61	67%
105	Stealth: Aurora Alpha	$0.0000	14.4s	65%	68	68	67	67	63	66%
114	Inception Mercury	$0.0074	14.4s	62%	69	68	65	63	61	65%
77.72%

Median	Evaluator	Top 3	Flop 3
91.9%	"Not X but Y" pattern overuse	100Gemma 3 4B 100ByteDance Seed 1.6 100GPT-5	2GPT-5 Nano 28Mistral Small 4 33Writer: Palmyra X5
47.0%	Adverb-first sentence starts	100LFM2 24B 95Writer: Palmyra X5 95Claude 3.5 Sonnet	0Qwen 2.5 72B 0Stealth: Aurora Alpha 0Inception Mercury
94.3%	Adverbs in dialogue tags	100o4 Mini 100Z.AI GLM 4.7 Flash 100Qwen 3.5 397B A17B	20GPT-4.1 Nano 31GPT-4.1 Mini 38Claude Haiku 4.5
86.8%	AI-ism adverb frequency	98ByteDance Seed 1.6 Flash 98ByteDance Seed 2.0 Lite 96GPT-4.1	60Gemma 3 4B 66Claude 3 Haiku 68GPT-4.1 Nano
100.0%	AI-ism character names	100Ministral 3 3B 100GPT-5 Nano 100Qwen 3.5 122B	80Z.AI GLM 5 Turbo 80Mistral NeMO 84Gemma 3 4B
100.0%	AI-ism location names	100Gemini 2.5 Flash Lite 100Gemini 2.5 Pro 100Qwen 3.5 Flash	96Claude Haiku 4.5
46.8%	AI-ism word frequency	84GPT-5.4 Mini (Reasoning) 84GPT-5.4 81GPT-5.4 (Reasoning)	3Inception Mercury 3GPT-4o Mini (temp=0) 3Claude 3.5 Haiku
100.0%	Cliché density	100Qwen 3.5 Plus (2026-02-15) 100Claude Opus 4.6 (Reasoning) 100Claude Sonnet 4.5	20GPT-4o, May 13th (temp=0) 33Mistral Small 3.2 24B 47Claude 3 Haiku
74.6%	Dialogue tag variety (said vs. fancy)	100GPT-5.4 Mini 100Claude Sonnet 4.5 100Claude Opus 4.6 (Reasoning)	0Qwen 3.5 Plus (2026-02-15) 8Gemini 3 Pro (Preview) 16Nemotron 3 Nano
16.3%	Em-dash & semicolon overuse	100Qwen 3.5 397B A17B 100GPT-5.4 Mini (Reasoning) 100GPT-4o, May 13th (temp=0)	0Claude Opus 4.6 0Qwen3 235B A22B Instruct 2507 0Claude Sonnet 4
100.0%	Emotion telling (show vs. tell)	100GPT-4o, Aug. 6th (temp=1) 100Qwen 3.5 27B 100Grok 4.20 (Beta, Reasoning)	88Mistral Small 3.2 24B 90GPT-4o Mini (temp=0) 93Hermes 3 70B
93.4%	Filter word density	100GPT-5.4 (Reasoning, Low) 100Ministral 3 14B 100GPT-5.4 (Reasoning)	8Inception Mercury 2 9Nemotron 3 Nano 11Stealth: Aurora Alpha
100.0%	Gibberish response detection	100Mistral Large 100Gemini 3 Flash (Preview) 100Mistral Small 4	41Llama 3.1 8B 87Rocinante 12B 98Qwen 2.5 72B
100.0%	Markdown formatting overuse	100Qwen 3.5 397B A17B 100Gemma 3 4B 100Z.AI GLM 5 Turbo	93ByteDance Seed 1.6 Flash
100.0%	Missing dialogue indicators (quotation marks)	100Stealth: Aurora Alpha 100Qwen 2.5 72B 100GPT-5 Mini	20Qwen 3.5 397B A17B 60Qwen 3.5 122B 60Gemini 3.1 Pro (Preview)
43.3%	Name drop frequency	94GPT-4.1 Nano 86Grok 4 Fast 84Nemotron 3 Nano	1Z.AI GLM 4.5 3GPT-5.4 Nano 3Claude 3.7 Sonnet
76.9%	Narrator intent-glossing	100GPT-4o, May 13th (temp=0) 100Qwen 3.5 Flash 100Gemini 3.1 Pro (Preview)	11GPT-5 Nano 17Nemotron 3 Nano 20GPT-4o, Aug. 6th (temp=1)
100.0%	Overuse of "that" (subordinate clause padding)	100Ministral 3 8B 100o4 Mini 100Gemma 3 4B	64Inception Mercury 74Mistral Small 3.2 24B 74Hermes 3 70B
100.0%	Paragraph length variance	100Grok 4.20 (Beta) 100Z.AI GLM 5 100Qwen 3.5 35B	53GPT-5 Nano 65ByteDance Seed 2.0 Lite 67Nemotron 3 Nano
90.4%	Passive voice overuse	100Grok 4.1 Fast 100Llama 3.1 8B 99Hermes 3 405B	65ByteDance Seed 2.0 Mini 70Nemotron 3 Super 76Nemotron 3 Nano
83.9%	Past progressive (was/were + -ing) overuse	100GPT-4o, May 13th (temp=1) 100DeepSeek-V2 Chat 100Nemotron 3 Super	0Gemini 3 Flash (Preview) 12Z.AI GLM 4.6 20Claude Sonnet 4.6 (Reasoning)
95.3%	Pronoun-first sentence starts	100DeepSeek V3 (2025-03-24) 100GPT-5.4 (Reasoning) 100Mistral Large 2	37Inception Mercury 40Qwen 3.5 27B 44Mistral NeMO
97.6%	Purple prose (modifier overload)	100Mistral Large 100ByteDance Seed 1.6 Flash 100o4 Mini High	84GPT-4.1 Nano 86Gemma 3 4B 88ByteDance Seed 2.0 Mini
100.0%	Repeated phrase echo	100Qwen 3.5 122B 100Gemini 3 Flash (Preview, Reasoning) 100Mistral Small Creative	—
100.0%	Sentence length variance	100Arcee AI: Trinity Large (Preview) 100GPT-5.4 Mini (Reasoning) 100GPT-5.4 Nano	99Cohere Command R+ (Aug. 2024) 99Hermes 3 70B 100Mistral Small 3.2 24B
58.3%	Sentence opener variety	95Rocinante 12B 94Claude 3.5 Haiku 91Claude 3.5 Sonnet	31Inception Mercury 32Stealth: Aurora Alpha 34GPT-5 Nano
42.4%	Subject-first sentence starts	93GPT-5.4 92Claude Sonnet 4.5 92Grok 4 Fast	0Stealth: Aurora Alpha 1Inception Mercury 2 2Inception Mercury
39.9%	Subordinate conjunction sentence starts	100Gemini 3.1 Flash Lite (Preview) 90Qwen 3.5 397B A17B 88Claude Sonnet 4.6	0Inception Mercury 0Claude Sonnet 4 0DeepSeek V3 (2025-03-24)
67.0%	Technical jargon density	100Rocinante 12B 100GPT-4o, May 13th (temp=0) 100Qwen 2.5 72B	0GPT-5 Nano 0Nemotron 3 Nano 0ByteDance Seed 2.0 Lite
33.8%	Useless dialogue additions	100GPT-5.4 (Reasoning, Low) 100Gemini 3.1 Flash Lite (Preview) 97Claude Sonnet 4.6	0Nemotron 3 Nano 0Gemma 3 4B 0Gemma 3 12B

Bad Writing Habits

Mystery: examining a crime scene

Performance Score Distribution (Top 20)

Price-Performance Score Distribution (Top 20)

Most Stable Models (Top 20)

Top Overall Models (Top 20)